Home // ALLSENSORS 2021, The Sixth International Conference on Advances in Sensors, Actuators, Metering and Sensing // View article

Combining Multiple Modalities with Perceiver in Imitation-based Urban Driving

Authors:
Shubham Juneja
Virginijus Marcinkevicius
Povilas Daniusis

Keywords: imitation learning; urban driving; perceiver

Abstract:
Traditional autonomous driving methods have relied on multiple sensor inputs for their success in decision making. Meanwhile, these methods require greater engineering effort as they consist of multiple modules than end-to-end methods which learn from data. In comparison, end-to-end methods rely on only a single modality and lack the ability to thoroughly generalise to new environments compared to traditional approaches. To enhance the current state-of-the-art methods, we propose using additional environmental information into an end-to-end learned method by employing the Perceiver architecture. The proposed technique aims to use more than one modality by fusing sensor data into a learner to generalise better in urban environments.

Pages: 43 to 44

Copyright: Copyright (c) IARIA, 2021

Publication date: July 18, 2021

Published in: conference

ISSN: 2519-836X

ISBN: 978-1-61208-875-4

Location: Nice, France

Dates: from July 18, 2021 to July 22, 2021